Warehouse Sync SQL v2 Reference

A SQL query can be provided as part of the Data Model request body with type set to sql. mParticle passes your SQL query to your warehouse, so be sure to use valid syntax for your warehouse. For example, each warehouse has slightly different ways to utilize SQL functions with different parameters.

Data Models and SQL Queries

The data in your warehouse is mapped to the user profile in mParticle according to the column names in the SQL you provide in the data model request following these rules:

• The provided SQL query supports being wrapped in an outer SELECT statement
• Column name casing follows the default identifier naming conventions for your specific warehouse, wrap column names in "" or [] to preserve casing.

• Every column is mapped to a user attribute except:

  • The column matches the optional load_timestamp_field_type field (case-insensitive) in the data model request
  • The column matches a reserved mParticle user or device identity column name

Reserved mParticle user or device identity column name

For a row to be associated to a profile, the following reserved column names are mapped to the mParticle User Identities or Device Identities. Additional columns are mapped as custom user attributes.

The following column names in your warehouse (case-insensitive) will be mapped to mParticle user identities automatically:

• mpid
• customerid,customer_id
• facebook
• twitter
• google
• microsoft
• yahoo
• email
• facebook_custom_audience_id
• other
• other_id_2
• other_id_3
• other_id_4
• other_id_5
• other_id_6
• other_id_7
• other_id_8
• other_id_9
• other_id_10
• mobile_number
• phone_number_2
• phone_number_3

The following column names in your warehouse (case-insensitive) will be mapped to mParticle device identities automatically:

• android_uuid
• ios_advertising_id
• push_token
• ios_idfv
• android_advertising_id
• amp_id
• roku_advertising_id
• roku_publisher_id
• microsoft_advertising_id
• microsoft_publisher_id
• fire_advertising_id
• mp_deviceid

Supported SQL Commands

The following SQL commands are fully supported:

• SELECT
• DISTINCT
• FROM
• ORDER BY
• GROUP BY
• WHERE
• JOIN
• AS
• ON
• IN

The following SQL commands are not supported:

• DELETE
• UPDATE
• DROP
• CREATE
• ALTER
• WITH

Example SQL Queries

Example 1: Import user’s propensity with a list of their favorite categories from your warehouse

mParticle’s demo database has a few user attributes and alongside a prediction for the likelihood they may purchase. The following query will import this alongside their favorite categories as a simple list and apply it to the corresponding Profile according to their customer_id.

SELECT
    a.date_updated,
    a.customer_id,
    a.firstname AS "$firstname",
    c.propensity_to_buy AS "propensity_to_buy",
    ARRAY_AGG(f.value) WITHIN GROUP (ORDER BY f.value ASC) AS "favorite_categories"
FROM mp.demo.attr a
JOIN mp.demo.calc c ON a.customer_id = c.customer_id
JOIN mp.demo.favs f ON a.customer_id = f.customer_id
GROUP BY 
    a.date_updated,
    a.customer_id,
    a.firstname,
    c.propensity_to_buy

Example 2: Import detail’s about a customer’s service tickets

mParticle’s demo ticket database contains details about the number of requests a user has made. The following query will import the number of open tickets to that user’s Profile according to their e-mail address.

SELECT
    u.email AS email,
    COUNT(t.id) AS "count_of_open_tickets"
FROM mp.demo_service.tickets t
JOIN mp.demo_service.users u
ON u.id = t.requester_user_id
WHERE t.status = 'open'

Troubleshooting SQL Queries

mParticle materializes your query by wrapping it in an outer SELECT statement to build more complex statements to execute against your data warehouse. These are the queries you will see looking at the history/audit logs in your data warehouse. For example, assume that a data model has the following query:

SELECT 
    a.scanned_timestamp_ms,
    c.propensity_to_buy
FROM demodw.demo.mp_dw_demo_attr a
JOIN demodw.demo.mp_dw_demo_calc c ON a.customer_id = c.customer_id

The query will be wrapped into (but not limited to) queries such as:

Query the number of rows in your provided data model. The values in the filter predicate are available as data_interval_start and data_interval_end in the pipeline run status API:

SELECT COUNT(*)
FROM
(
SELECT a.scanned_timestamp_ms, c.propensity_to_buy
FROM demodw.demo.mp_dw_demo_attr a
JOIN demodw.demo.mp_dw_demo_calc c ON a.customer_id = c.customer_id
)
WHERE SCANNED_TIMESTAMP_MS BETWEEN '2023-03-01 14:28:55+0000' AND '2023-03-01 14:41:17+0000'

Query the number of columns in your provided data model:

SELECT *
FROM
(
SELECT a.scanned_timestamp_ms, c.propensity_to_buy
FROM demodw.demo.mp_dw_demo_attr a
JOIN demodw.demo.mp_dw_demo_calc c ON a.customer_id = c.customer_id
)
LIMIT 0

Query generated from a scheduled sync run. The values in the filter predicate are available as data_interval_start and data_interval_end in the pipeline run status API.

SELECT OBJECT_CONSTRUCT_KEEP_NULL(*) 
FROM
(
SELECT a.scanned_timestamp_ms, c.propensity_to_buy
FROM demodw.demo.mp_dw_demo_attr a
JOIN demodw.demo.mp_dw_demo_calc c ON a.customer_id = c.customer_id
)
WHERE SCANNED_TIMESTAMP_MS BETWEEN '2023-03-01 14:28:55+0000' AND '2023-03-01 14:41:17+0000'